Fault Detection Of Networked Control System Based On Kalman Filter

With the development of the network technology and computer technology, the control system based on network has substituted traditional point-to-point control system. The closed-loop control system through network is called networked control systems (NCS). The NCS can realize resources sharing and improve the system reliability and flexibility; it can be installed and maintained easily.Fault detection has increasingly become a very important research filed. In practical control process, some factors such as the change of system parameter, disturbance and noise will affect residual signal, so it is essential to treat them as unknown input and isolate above factors from residual signal. The process is called Fault detection which obtained attention by scholars.The fault detection of NCS is a new research topic. The idea based on observer is the fault is generated by residual signal. The design of the residue generator is one of the core problems of fault detection of networked control system. Namely, on the one hand, the designed residue generator could maximize the sensitivity of residual to fault and simultaneously guarantee the robustness of the residual signal to the reference input, disturbance signal as well unknown input.The paper considers the NCS time-delay, analyze basic question of NCS and expatiate some common observers in detail. Then analyze the characteristics of NCS based on observer method from two aspects. On the one hand, fault detection, only considering the time-delay between sensor and controller, an optimal estimator with time-delay compensation function is constructed to estimate the system whole state, at last the conclusion that the system residual signal is only related to fault and noise is obtained. On the other hand, fault-tolerant control, the sensors failing to transmit data are treated to be invalidated temporarily, with the theory of the fault-tolerant control to design the controller. The controller is structured by two portion, one is state observer which reconstructs whole state according to measured output; another is control law which feedback reconstructing whole state directly, at last, the system is stable.